Yuntao Ma / ETH Zurich
We’ve already watched robotic dogs carry gear up steep trails and even assist firefighters. Now, researchers at ETH Zurich in Switzerland are testing their four-legged machines on the badminton court, training them to play at a level comparable to a young child.
A New Control System for ANYmal
The team developed a new control system for the institute’s ANYmal robot that relies on two cameras to track the shuttlecock in flight, calculate its path, and position the robot to return the shot with a racket mounted on a multi-axis arm. You can see ANYmal in action in the video below.
What makes this especially interesting is the use of reinforcement learning. The program runs countless virtual simulations, teaching the control system how to move efficiently and how to swing the racket arm to return incoming shots.
Precision and Balance in Motion
Thanks to this setup, the robot can intercept the shuttlecock with precision, while moving quickly and maintaining its balance. Achieving this requires seamless coordination between its vision-based perception and its motor skills.
Although the reinforcement-learning control system is new, ANYmal itself has been in development for years and is already commercially available for autonomous inspection in hazardous environments, such as detecting gas leaks in power plants. Its agility has improved significantly in recent updates.
ANYbotics
Previously, the robot mastered rough hiking paths and even learned parkour, which gave it the ability to balance on uneven ground—an asset in potential disaster-response missions among collapsed structures.
With its latest badminton-playing upgrade, ANYmal could one day support athletic training. However, with a price tag of around US$150,000, it’s more likely to show up first in elite sports facilities once the technology matures.
Read the original article on: New Atlas
Read more: Robotics Combined with Virtual Reality Allows Seamless, Natural Engagement